The Acute Box cis-Element in Human Heavy Ferritin mRNA 5′-Untranslated Region Is a Unique Translation Enhancer That Binds Poly(C)-binding Proteins*

Intracellular levels of the light (L) and heavy (H) ferritin subunits are regulated by iron at the level of message translation via a modulated interaction between the iron regulatory proteins (IRP1 and IRP2) and a 5′-untranslated region. Iron-responsive element (IRE). Here we show that iron and interleukin-1β (IL-1β) act synergistically to increase H- and L-ferritin expression in hepatoma cells. A GC-rich cis-element, the acute box (AB), located downstream of the IRE in the H-ferritin mRNA 5′-untranslated region, conferred a substantial increase in basal and IL-1β-stimulated translation over a similar time course to the induction of endogenous ferritin. A scrambled version of the AB was unresponsive to IL-1. Targeted mutation of the AB altered translation; reverse orientation and a deletion of the AB abolished the wild-type stem-loop structure and abrogated translational enhancement, whereas a conservative structural mutant had little effect. Labeled AB transcripts formed specific complexes with hepatoma cell extracts that contained the poly(C)-binding proteins, iso-αCP1 and -αCP2, which have well defined roles as translation regulators. Iron influx increased the association of αCP1 with ferritin mRNA and decreased the αCP2-ferritin mRNA interaction, whereas IL-1β reduced the association of αCP1 and αCP2 with H-ferritin mRNA. In summary, the H-ferritin mRNA AB is a key cis-acting translation enhancer that augments H-subunit expression in Hep3B and HepG2 hepatoma cells, in concert with the IRE. The regulated association of H-ferritin mRNA with the poly(C)-binding proteins suggests a novel role for these proteins in ferritin translation and iron homeostasis in human liver.

The mechanisms governing the regulation of ferritin mRNA translation are complex, but their elucidation is critical to understanding iron homeostasis. Iron and oxidative stress are known to modulate the first stage of translation of ferritin mRNAs when the 43 S ribosome subunit attaches to the 5Ј cap-specific M 7 GpppN in the 5Ј-UTR 1 of L-and H-ferritin mRNAs (1)(2)(3)(4). The iron regulatory proteins (IRP1 and IRP2, iso-IRPs) play a central role in regulating ferritin mRNA translation. During conditions of intracellular iron chelation with desferrioxamine (DesF) and oxidative stress, the IRPs bind with higher affinity to the conserved iron-response element (IRE) RNA stem loop 40 nucleotides (nt) downstream of the 5Ј cap sites of the L-and H-ferritin mRNAs (1). This translational repression event prevents attachment of the small ribosome subunit to the 5Ј cap sites of the L-and H-ferritin mRNAs and inhibits ferritin translation (2). In contrast, after iron influx the iso-IRPs are released from the 5Ј cap IREs, and ferritin translation is no longer inhibited, increasing the cellular iron storage capacity (2,3). The IRP2 knock-out mouse, which is characterized by unregulated ferritin mRNA translation and ferritin accumulation in neurons and gut epithelial cells in a gene-dose manner, validated these observations in vivo (4). Recently zinc and cadmium were reported to interfere with the RNA binding activity of IRP-1, extending the spectrum of IRP binding modulators to these two metal elements (5).
Thyroid hormone (T 3 ), which displaces iso-IRPs from the iso-IREs in iron-loaded cells, increases ferritin translation (6). Other factors also regulate ferritin expression via altered IRP-IRE interactions. These include phorbol esters, endothelial growth factor and thyrotropin-releasing hormones, each of which modulates the phosphorylation status of the iso-IRPs (7). The thyrotropin-releasing hormone/endothelial growth factorinduced changes in ferritin subunit synthesis were IRP-dependent in one pituitary cell line and IRP-independent in another, suggesting that other sequences within L-and H-ferritin mRNAs contribute to translation regulation (2). IRP-independent ferritin subunit synthesis is also induced in human epidermal A431 cells when they are infected with Neisseria meningitidis (8). These data indicate that mechanisms other than the IRP-IRE interaction can modulate L-and H-ferritin mRNA translation.
Interleukin-1 (IL-1) appears to control the rate of L-and H-ferritin subunit synthesis at the second stage of 43 S ribosome translation scanning, immediately upstream from the start codon before the complete 80 S ribosome translates the open reading frame into protein (9 -11). We have shown previously that IL-1␤ stimulates the rate of L-and H-ferritin subunit translation in both hepatomas (11). In particular, IL-1␤ induced both L-and H-ferritin mRNAs and activated their recruitment to the polyribosome from stored ribonucleoproteins (11). A distinct 63-nt GϩC-rich RNA sequence 105 nt downstream from the H-ferritin mRNA IRE was found to confer 2-3-fold IL-1␤-dependent enhancement of the translation of hybrid H-ferritin-chloramphenicol acetyltransferase (CAT) reporter mRNAs in human HepG2 hepatoma cells (13). This IL-1␤-dependent translation enhancement element encodes a core 25-nt consensus motif (CGCCGCGCAGCCACCGCCGC-CGCCG, the acute box (AB)), homologous to sequences encoded in the 5Ј-UTRs of several hepatic acute phase reactant mRNAs, including ␣-1 acid glycoprotein (AGP), ␣-1 antitrypsin (␣ 1 AT), and haptoglobin (14,15). In both hepatoma and endothelial cells, a highly homologous L-ferritin AB also conferred translation enhancement to reporter mRNA (16). A third transcript containing the AB is the Alzheimer amyloid precursor protein (APP), which confers IL-1␤-induced regulation of APP translation (9). Based on the fact that repression of upstream IRE-dependent translation by DesF was dominant over the IL-1␤induced stimulation of ferritin translation (9), we suggested previously that the AB operated to enhance the 60 S ribosome joining step of ribosome translational scanning of H-ferritin mRNA by the 43 S ribosome according to the Kozak model (10) rather than providing an internal ribosome entry site (IRES).
Current models for ferritin translation only incorporate irondependent IRP-1 and IRP-2 binding to the 5Ј-UTR of H-ferritin mRNA for modulating the interaction with the incoming ribosome and translation initiation factors (12). There are currently no reports identifying other RNA-binding proteins that may selectively interact with the H-ferritin mRNA 5Ј-UTR via the AB to regulate basal ferritin translation. The poly(C)-binding proteins (PCBPs) are a structurally diverse family, including heterogeneous nuclear ribonucleoprotein K and ␣CP1-4 (17). They bind mRNA sequences that contain either a single C run (heterogeneous nuclear ribonucleoprotein K) or stretch of C's (␣CPs) via their K-homology (KH) domains (18). The PCBPs have been implicated in regulation of mRNA stability, translation silencing (mainly through interactions with 3Ј-UTR sequences), and enhancement. The PCBP family members bind to each other and to other mRNA binding proteins, including AUF-1 (AU-rich element RNA-binding protein) and PABP, which are important in modulating decay of globin mRNAs (19). ␣CP2 regulates translation of poliovirus mRNA via a specific IRES (20,21) and 15 lipoxygenase mRNA translation via a C-rich element in the 3Ј-UTR (17,22). Pertinent to this report, human HepG2 and Hep3B hepatoma cells are reported to contain ␣CP1 and ␣CP2 that bind 3Ј-UTR elements of erythropoietin and tyrosine hydroxylase mRNAs (23).
To determine the structural and functional characteristics of the AB in the regulation of ferritin translation, we tested the effects of mutations of the AB element on IL-1-and iron-dependent translation in two different human liver cell lines. We found that the predicted shape and sequence of the AB ele-ment, independent of the IRE, was critical for maintaining enhanced base-line translation in addition to a IL-1␤-induced increase in translation. In a time course the AB responded to IL-1␤ signals at the same time that endogenous ferritin was induced. Transfections using an H-ferritin promoter and 5Ј-UTR sequences together with either the wild-type AB or an equivalent length scrambled AB demonstrated that the AB was a novel basal and IL-1-dependent translational enhancer. Furthermore, in RNA gel-shift (RNA electrophoretic mobility shift assay (REMSA)) and UV cross-linking experiments we demonstrated that the H-ferritin mRNA AB cis-element interacts specifically with recombinant ␣CP1 and that both ␣CP1 and ␣CP2 associate with H-ferritin mRNA in vivo. These data provide evidence that the AB is an important contributor to Hferritin expression in human liver cells, responsible for enhancing both basal and IL-1-mediated translation and also identified the PCBPs (␣CP1 and ␣CP2) as novel H-ferritin mRNA-binding proteins that may act in a coordinate manner with the IRPs to control the overall rate of ferritin translation.
Plasmid Constructs for Transient Transfections-pUC-HFER (a gift from Dr. J. Drysdale, Tufts University School of Medicine, Boston, MA) containing a 454-bp SstI fragment from the H-ferritin gene cloned into pUC12 (9, 13) comprises 162 bp of H-ferritin gene sequence upstream of the 5Ј-UTR cap site and 292 bp of the first exon including 5Ј leader sequences. All subsequent constructs were derived from H-ferritin sequences encoded by pUC-HFER, and their identity was confirmed by double-stranded DNA sequencing to preclude the presence of artificial AUG sites upstream of the CAT initiation codon. HIRECAT contains 302 nt. A SstI-StyI fragment from pUC-HFER ligated into the unique SstI and HincII sites in the polylinker of pUC12CAT (Fig. 1C), directly in front of the CAT start codon. 5Ј-UTRCAT contains 363 nt. A SstI-DdeI fragment from pUC-HFER ligated into the unique SstI and HincII sites in the polylinker of pUC12CAT (a gift from Dr. W. Chin, Harvard Medical School, Boston, MA) (Fig. 1D). Each transcript derives from the bona fide H-ferritin core promoter. The AB-HIRECAT and Scr-HIRE-CAT constructs were prepared by inserting annealed 63-base oligonucleotide cassettes (acute box (AB) or scrambled (Scr) sequences) into PstI-digested HIRECAT (Fig. 3C). Oligonucleotides were designed to encode a PstI site at the 5Ј and 3Ј ends (24,25). After annealing, oligos were digested and ligated into PstI-digested HIRECAT. All constructs were sequenced to confirm the correct alignment and incorporation of regulatory domains.
To prepare pSV2CAT-derived constructs, pSV2(Ac)CAT and pSV2(rev)CAT were prepared by ligating the 63 bp of StyI-DdeI fragment from the 5Ј-UTR of the H-ferritin mRNA (previously 5Ј end filled by Klenow polymerase) (Fig. 1, A and B) into a unique StuI site of pSV2CAT, residing 42 nt downstream of the SV40 early T-antigen promoter and 43 nt upstream of the CAT gene start codon (Fig. 1B). pSV2(Mc)CAT and pSV2(⌬3)CAT contain PCR-mutated versions of the 63-bp StyI-DdeI fragment from the 5Ј-UTR of the H-ferritin mRNA inserted in the StuI site of pSV2CAT (Fig. 1, A and B). pRSVLuc is as described (9).
CAT and Luciferase (Luc) Enzyme Assays-The level of CAT expression was determined in cells that had been stimulated with iron and/or IL-1␤ at various times. CAT assays were performed as described (13) on lysates from HepG2 and Hep3B transient transfections using a liquid scintillation counting assay. To validate the data, CAT assays were also performed with thin layer chromatography (13) and quantified using a Betagen scanner (Betagen Corp., Waltham, MA). As a transfection control, each cell lysate (150 l) was also assayed in duplicate for luciferase activity using a luminometer and luciferin (Analytical Luminescence Laboratory, San Diego, CA) according to the manufacturer's instructions. CAT enzyme-linked immunosorbent assay assays were also performed according to manufacturer's instructions (Roche Applied Science). Briefly, cells were lysed in Roche reporter gene lysis buffer after washing in phosphatebuffered saline. Cleared lysate (200 l) was incubated in anti-CAT microplates (37°C, 60 min) followed by the addition of anti-CATdigoxigenin (200 l, 37°C, 60 min) and anti-digoxigenin-peroxidase (200 l, 37°C, 60 min). Samples were analyzed by absorbance at 405 nm, where CAT concentration was determined against a known concentration CAT standard curve and normalized with respect to sample protein concentration determined by Bio-Rad protein assay.
Ferritin and CAT mRNA Detection by Slot-blotting and RNase Protection-Transfected HepG2 and Hep3B cells (10 7 ) were treated with IL-1␤ (1 ng/ml), Fe 2 Tf (5 M) or DesF (100 M), RNA extracted using the guanidinium-HCl purification method (9,11), DNase I (10 g/ml) treated for 30 min at 37°C, and slot-blotting was performed after denaturing the RNA (20 g) for 15 min at 65°C in formaldehyde solution (6.15 M formaldehyde, 0.15 M sodium citrate, 1.5 M NaCl). RNA was applied to slots aligned on nitrocellulose membranes. A 1.635kilobase HindIII-BamHI fragment coding for the CAT gene from pSV2CAT was random-primed-labeled (specific activity ϭ 2 ϫ 10 8 cpm/ g) and used as hybridization probes. Overnight hybridization and washing conditions of the filters was as described previously (9). The CAT gene probe, which specifically hybridized to each RNA slot, was quantified by either (i) directly counting each excised slot in a scintillation counter or (ii) comparative densitometry with known denatured pSV2CAT DNA standards (0.01-1 ng) (2). RNase protection assays were performed after T7 polymerase transcription of a 261 nt. 32 P-Labeled cRNA from HindIII digested DNA isolated from the CAT gene subclone, pBSCAT. The 32 P-labeled CAT cRNA (1 ϫ 10 8 cpm) was hybridized with total HepG2 RNA (20 g) in hybridization buffer (80% formamide, 40 mM Pipes. pH 6.7, 0.4 M NaCl, 1 mM EDTA) for 16 h at 45°C. RNase A (40 g/ml) and RNase T1 (2 g/ml) were added, and protected cRNAs and kinase 32 P-labeled HaeIII-digested X174 DNA standards were separated by denaturing PAGE (6% polyacrylamide, 7 M urea gel).
Cytoplasmic Protein Extracts-Cytoplasmic protein extracts of treated HepG2 cell cultures were prepared as described (7). Briefly, cells were scraped from culture dishes in chilled phosphate-buffered saline, centrifuged at 450 ϫ g for 4 min at 4°C, washed again in phosphate-buffered saline, and then incubated for 20 min in cold cytoplasmic extraction buffer (10 mM HEPES, 3 M MgCl 2 , 40 mM KCl, 5% glycerol, 0.2% Nonidet P-40, 1 mM dithiothreitol) containing freshly added protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 2 g/ml aprotinin). Lysates were cleared by centrifugation at 4°C for 10 min at 12,100 ϫ g, and the supernatant was snap-frozen in liquid nitrogen and stored at Ϫ80°C. Protein concentra- tions were determined by Bio-Rad protein assay kit.
IP-RT-PCR Assay-HepG2 cells were grown (ϮFe 2 Tf (10 M) and ϮIL-1␤ (1 ng/ml)), and cytoplasmic extracts were harvested in cytoplasmic extraction buffer as described (7). Cytoplasmic extract (200 g) was added to 10 g of ␣CP1 or ␣CP2 antibody (Ab). After incubation on ice for 45 min, 5 g each of protein A (Amersham Biosciences) and protein G (Sigma) beads were added to all samples, which were mixed for a further 45 min at 4°C. The samples were centrifuged at 2000 ϫ g for 2 min, and the supernatants were removed for RNA extraction. The pelleted beads were washed with cold cytoplasmic extraction buffer (10 ϫ 1 ml), and RNA was extracted using Trizol reagent (Promega). Reverse transcription was performed with oligo(dT) 25 and standard procedures. PCR was performed for 35 cycles (denaturation at 95°C/30 s, annealing at 55°C/45 s, extension at 72°C/60 s) with the H-ferritincoding region specific primers, FerIP-sense (5Ј-ATG ACG ACC GCG TCC-3Ј) and FerIP-antisense (5Ј-TTG CAT TCA GCC CGC-3Ј), designed to produce an amplicon of 290 bp. PCR products were resolved on an ethidium bromide-stained 2% agarose gel. IP-RT-PCR was also performed on samples in which HepG2 total RNA (protein free) was incubated with ␣CP1 and ␣CP2 Abs to ensure cell protein was required for RNA co-immunoprecipitation.

Iron Is Required for IL-1␤ Induction of L-and H-ferritin
Subunit Synthesis-To investigate the relative impact of IL-1␤ and iron (Fe 2 Tf) on the synthesis of ferritin subunits in vivo, we incubated HepG2 cells with each ligand over a time course of 2-6 h. Synthesis of L-and H-ferritin subunits was increased by ϳ4-fold by IL-1␤ relative to untreated counterparts at 6 h (Fig.  2, A and D). A combination of Fe 2 Tf and IL-1␤ increased ferritin synthesis to ϳ20-fold above control (Fig. 2, A and D), suggesting a synergistic effect of the two ligands when combined. The synthesis of ␣ 1 AT, another liver protein remained unchanged (see These data are consistent with other reports that demonstrate ferritin synthesis is induced by IL-1␤ up to 16 h post-induction (9,11,13,16). To ensure that the changes in ferritin synthesis were translational, steady state levels of H-and L-ferritin mRNAs were examined and shown to remain unchanged (data not shown and as previously reported (16)).
In other experiments Fe 2 Tf commenced induction of ferritin synthesis (both L and H subunits) within 2 h, whereas there was little effect of IL-1␤ until 6 h post-treatment, when there was a marked synergistic effect (Fig. 2, C and E). The effect of an overnight incubation with a transferrin receptor antagonist antibody (TfR Ab, a gift of Paul Seligman, University of Denver, CO) on the induction of ferritin synthesis by Fe 2 Tf and IL-1␤ was investigated. Preincubation with TfR Ab did not affect the ability of IL-1␤ to stimulate ferritin translation (Fig.  2F). However, the TfR Ab did effectively block TfR-mediated iron uptake as measured by 59 Fe-transferrin uptake (data not shown) and abrogated Fe 2 Tf-induced ferritin translation (Fig.  2F). Thus, IL-1␤-dependent ferritin translation was maintained despite lowered levels of intracellular iron (in the presence of inactivating TfR Ab). In contrast, complete chelation of intracellular iron pools with DesF had a profound effect on the IL-1␤ induction (Fig. 2A, lane 5). We conclude that permissive levels of iron are required for IL-1␤-induced up-regulation of ferritin translation.

Contribution of the AB Element to Ferritin Translation in HepG2 Cells Stimulated with Iron and IL-1␤; the AB Confers
Enhanced Base-line Translational Enhancement-To determine the relative contribution of the AB versus the IRE to the ferritin translational enhancement described above, we generated constructs of the 5Ј-UTR of the H-ferritin gene (HIRECAT and 5Ј-UTRCAT) for transfections. The HIRECAT construct harbors a CAT reporter gene driven by the natural H-ferritin promoter and encodes 144 nt of the H-ferritin 5Ј-UTR, containing an IRE stem-loop 30 nt from the 5Ј cap site but lacking the IL-1␤-responsive AB sequence (ϩ145-208) (Fig. 1C). The 5Ј-UTR-CAT transfectants encode the full-length H-ferritin mRNA 5Ј-UTR driven by the natural H-ferritin gene promoter and contain both the IRE stem loop and the AB domain (Fig. 1D).
When HepG2 cells were transfected with 5Ј-UTRCAT and treated with either IL-1␣ or IL-1␤, we found that there was no increase in CAT translation at 2 h (Fig. 3A). However, at 6 -24 h both basal and IL-1-stimulated CAT translation increased severalfold (ϳ4-fold for IL-1␣ and ϳ3-fold IL-1␤) (Fig. 3A). The data in these experiments closely paralleled the time course we had observed for IL-1␤ induction of endogenous H-ferritin expression in the same cells (Fig. 2E).
Iron induced an 8-fold increase in CAT gene expression in HIRECAT transfectants (Fig. 3B), but IL-1␤ had no effect. These data confirmed that a functional IRE was not sufficient to confer IL-1␤-dependent translation to a CAT reporter mRNA. As before, IL-1␤ conferred a 2.6-fold increase in CAT gene expression in the 5Ј-UTRCAT transfectants (Fig. 3B).
Interestingly, the fold induction of iron-dependent translation was lower for 5Ј-UTRCAT construct (3-fold) than for HIRECAT (8-fold) when compared with untreated control. This maybe FIG. 3-continued because the base-line CAT expression was higher for the 5Ј-UTRCAT transfectants relative to the HIRECAT transfectants (ϳ2-2.5-fold, n ϭ 3), such that the absolute change induced by iron was substantially smaller. The expected greater maximal induction of 5Ј-UTRCAT expression under conditions of iron influx was not observed as seen with endogenous ferritin levels (Fig. 2), suggesting that other regions of the ferritin mRNA maybe involved in translational regulation (e.g. 3Ј-UTR) (26,27).
A Scr 63-nucleotide spacer was generated (Fig. 3C) to directly investigate if the presence of the wild-type AB was required to confer basal and IL-1-dependent translational enhancement by H-ferritin 5Ј-UTR sequences. In transfection studies using HepG2 cells, we consistently observed that chimeric H-ferritin-CAT mRNA derived from AB-HIRECAT transfectants was translated 3-fold more efficiently than Hferritin-CAT mRNA derived from Scr-HIRECAT transfectants (Fig. 3D). In AB-HIRECAT transfectants, IL-1␣, IL-1␤, and iron induced CAT reporter translation (2-fold relative to untreated (n ϭ 6)), whereas only iron remained active in Scr-HIRECAT counterparts. We concluded that the presence of the wild-type AB upstream of the CAT start codon increased reporter translation, whereas an equivalently located scrambled insert was unable to confer either increased base-line or IL-1dependent translation relative to HIRECAT. These data indicated that the AB was active as a sequence and/or structurespecific translation enhancer element (compare Fig. 3, E  and F).

The H-ferritin AB Domain Is an Autonomous Base-line Translation Enhancer That Also Mediates IL-1␤-induced
Translation-The HIRECAT and 5Ј-UTRCAT based transfections described above provided the first evidence that the AB element in the H-ferritin mRNA conferred increased base-line translation in addition to being an IL-1␤-responsive translational enhancer. To further characterize the H-ferritin AB-mediated translation regulation, we generated a series of constructs for comparative analysis. The wild-type Ac transcript (pSV2(Ac)CAT) harbors a chimeric 137-nt 5Ј-UTR consisting of viral sequences with the AB inserted in the sense orientation upstream of the CAT reporter start codon, whereas rev (pSV2(rev)CAT) was used for an antisense comparison. These constructs encode the same length 5Ј-UTR (137 nt) and transcribe a hybrid CAT gene under the translational control of sense (Ac) and antisense (rev) acute box sequences (63 nt). In HepG2 cells, the Ac construct induced a 2.2-fold increase in basal CAT expression compared with the parental vector, pSV2CAT (Fig.  4A, n ϭ 9). IL-1␤ induced a reproducible increase in CAT expression by 75% in cells transfected with the wild-type Ac construct (Fig. 4A, n ϭ 6). No significant change in CAT expression was observed with the IL-1␤-treated antisense (rev) transfectants or the control transfectants (pSV2CAT) (Fig. 4A). Of particular interest, the antisense construct also decreased basal expression relative to control (Fig. 4A) despite the fact that the predicted 2°s tructure (28) for the Ac and rev stem-loops is similar (with a ⌬G ϭ Ϫ40 kcal/mol). RNase protection assay confirmed no change in CAT mRNA levels with IL-1␤ treatment, indicating that the effect observed was at the level of translation (Fig. 4B). Taken together these data indicate that the H-ferritin mRNA AB is a significant translational enhancer that not only confers IL-1␤-dependent CAT expression but also acts as a powerful baseline translational enhancer.
Originally the pSV2(rev)CAT construct was employed to exclude the possibility that the AB was a transcriptional enhancer. Gillies et al. (29) have shown that transcriptional enhancer elements operate in an orientation-independent fashion. Our data consistently demonstrate that the AB ele-ment is active only when inserted in front of a heterologous start codon in the sense orientation (as in the pSV2(Ac)CAT construct), a finding consistent with the presence of novel translational enhancer activity.
To further investigate the function of the AB on translational efficiency, HepG2 cells were transfected with either pSV2CAT or pSV2(Ac)CAT, and the CAT activity was standardized to CAT mRNA (as measured by slot-blotted CAT mRNA levels) (16). In these experiments, the H-ferritin AB increased CAT expression by 2.8-fold, as shown by the change in the slope of kinetic induction for pSV2CAT relative to pSV2(Ac)CAT gene expression (Fig.  4C). In other experiments we determined the maximal induction of translation mediated by the AB from the reaction slopes of graphs of CAT expression from cells transfected with either the Ac or rev or parental vector. We found that pSV2(Ac)CAT lysates accumulated 3.5-fold more CAT than pSV2(rev)CAT and that pSV2(Ac)CAT lysates accumulated CAT 2.2-fold faster than the parental pSV2CAT lysates (data not shown).

Translation Enhancement by the H-ferritin mRNA AB in a
Second Hepatoma Cell Line (Hep3B Cells)-To investigate the cell line specificity of the translation enhancement of the Hferritin AB, Hep3B hepatoma cells were transfected with pSV2(Ac)CAT and pSV2CAT. Similar to HepG2 cells, the AB increased CAT expression by ϳ3-fold (corrected against CAT mRNA levels using slot-blot analysis as above) (Fig. 5A).
A marked dose-and time-dependent induction of CAT expression was evident when cells transfected with pSV2(Ac)CAT were treated with IL-1␤ (ϳ4-fold increase in response, Fig. 5, B and C). In contrast, no significant change in CAT activity was observed with the pSV2CAT transfectants (Fig. 5, B and C). These data confirm that the effects of the AB in these two human liver cell lines are very similar, although there is a consistently larger effect on translation in the Hep3B cells. The greater induction by IL-1␤ of pSV2(Ac)CAT in Hep3B cells may reflect the fact that base-line translation is lower in Hep3B cells compared with HepG2 cells, and greater induction of mRNA translation in Hep3B cells is required to achieve the same final endpoint ferritin levels as in HepG2 cells.
To verify the importance of the IL-1␤ pathway in mediating the AB-mediated translation enhancement, an IL-1␤-neutralizing Ab (Invitrogen) was added. The typical IL-1␤-induced increase in CAT expression (ϳ2.5-4-fold) was markedly reduced in the presence of IL-1␤-inactivating Ab (Fig. 5D). Inclusion of normal rabbit serum in all samples had no influence in the pattern of induction (data not shown). These data are consistent with a role for IL-1␤ receptor signaling in mediating translation of H-ferritin mRNA in Hep3B cells via the AB element.
Effect of Mutations and Deletions within the H-ferritin mRNA AB Sequence on Translation-To investigate the effects of mutations within the AB region on translational efficiency, we generated two key mutations of the wild-type H-ferritin sequence: (i) pSV2(⌬3)CAT (⌬3) and (ii) pSV2(Mc)CAT (Mc) (Fig. 1, A-B). The ⌬3 vector encodes a CAC deletion (⌬173-175) from the loop region of the AB stem-loop and is predicted to substantially alter folding of the AB domain (Supplemental Fig. 1C). The Mc construct contains a C 175 to A 175 mutation in the loop that maintains correct folding of the AB (Supplemental Fig. 1B). In transfected HepG2 cells, base-line CAT activity of ⌬3 was reduced by 6 -7-fold relative to Ac and reduced 2-fold relative to parental vector (Fig. 6A). IL-1␣ and IL-1␤ increased CAT gene expression in the Mc transfectants (Fig. 6A), similar to that observed in Ac transfectants. We also investigated the effects of IL-6 on CAT activity. Multiple experiments showed that IL-6 caused a modest reduction (10 -20%) in Ac-CAT and Mc-CAT translation in pSV2(Ac)CAT and pSV2(Mc)CAT transfections (see Fig. 6A, in which n ϭ 8, and Fig. 6B, where n ϭ 5).
Base-line CAT expression was consistently 2-fold higher in Mc transfectants than in ⌬3 transfectants and Ͼ30% above that of the parental vector (Vec). These data illustrated the ligand specificity of IL-1␣/␤ versus IL-6 and the importance of maintenance of the AB stem-loop structure and sequence for preservation of the translational enhancement. Interestingly, the AB antisense construct pSV2(rev)CAT and ⌬3 each abrogated base-line and IL-1␤/IL-6-treated CAT expression similarly (Fig. 6B).
We next examined all of the constructs together in a time course experiment from representative lysates (n ϭ 4) taken from the complete set of HepG2 transfectants. The Ac and Mc constructs generated 3-4-fold more CAT gene expression than pSV2CAT (Fig. 6C). Both ⌬3 and rev transfectants exhibited 3-4-fold less CAT activity after normalization than pSV2CAT alone (Fig. 6C). The rank order for the available AB elements as translational regulators of downstream CAT reporter expression was Ac ϭ Mc Ͼ Vec Ͼ ⌬3 ϭ rev. Taken together, these data illustrate the potency of the H-ferritin AB element as a

FIG. 6. Wild-type configuration of the H-ferritin mRNA AB enhancer is required to maintain both IL-1␤dependent-and base-line translation enhancement. Panel A,
HepG2 cells were transfected with 10 g of either pSV2(Ac)CAT (wild-type, Ac), pSV2(⌬3)CAT (⌬3), pSV2(Mc)CAT (Mc), or pSV2CAT (Vec) and treated with and without IL-1␣, IL-1␤, or IL-6 at 1 ng/ml. The CAT activity was determined as above and expressed as CAT activity per ng of CAT mRNA. Error bars, S.D.; n ϭ 4. Panel B, HepG2 cells were transfected with 10 g of either pSV2(Ac)CAT (Ac), pSV2(rev)CAT (rev), or pSV2(⌬3)CAT (⌬3) and treated with and without either IL-1␤ or IL-6 (1 ng/ml). The CAT activity was assessed by the percent [ 14 C]CAP (chloramphenicol) acetylated using TLC (13). n ϭ 4. Panel C, HepG2 cells were transfected with 10 g of Ac, ⌬3, Mc, rev, or Vec together with pRSVLuc (5 g) (as control), and CAT was activity determined over a 100-min time course as above. CAT activity ( 3 H-labeled acetylated chloramphenicol) was standardized against luciferase activity. Panel D, predicted 2°structure of the H-ferritin AB stem-loop (33 nt) (28). Panel E, RNase T1 digestion of H-ferritin AB RNA (33 nt). RNA fragments of undigested-and RNase T1-digested 32 P-labeled H-ferritin AB cRNAs were resolved by denaturing PAGE (20% sequencing, 7 M urea gels). base-line translational regulatory domain and also suggest that preservation of the stem-loop structure is essential for both base-line and IL-1␤-induced translation enhancement by the H-ferritin mRNA AB. That the AB C-A mutation (Mc) decreased base-line CAT activity relative to wild-type AB suggests that there is also some sequence specificity for base-line translation enhancement. This may correlate with a role for altered affinity for RNA-binding proteins between the Mc mutant and wild-type AB.
Given these transfection data, we further examined the predicted 2°structure of the wild-type H-ferritin AB region (Fig.  6D). To obtain direct physical data concerning folding of the H-ferritin AB mRNA, we labeled the 33-nt H-ferritin AB cRNA for use in RNase T1 assays. RNase T1 cuts RNA specifically at the 3Ј end of guanosine (G) residues and adjacent nucleotides in single-stranded RNA, GpN, and we predicted that it could potentially cleave at eight GpN sites in the probe. Interestingly, however, RNase T1 digestion demonstrated a single major product at M r ϳ 32 nt (Fig. 6E, n ϭ 3), indicating that the 33-nt H-ferritin AB was highly resistant to RNase T1 degradation. This suggests that the H-ferritin cRNA probe folded into a higher order structure that is cleaved only once at the GC bond at the 5Ј end to generate the 32-nt cRNA. After thermal denaturation at 95°C for 10 min, the AB RNA structure was completely digested (data not shown).

The H-ferritin mRNA AB Is a Target for Proteins from HepG2
Cells, Including the Poly(C)-binding Proteins 1 and 2 (␣CP1 and ␣CP2)-Analysis of the AB and surrounding sequences permitted folding of the H-ferritin AB element into a stable stem-loop, with the loop comprising the sequence CCACCG, where the terminal CG nucleotides are base-paired (Supplemental Fig. 1A). A single-stranded 5-nt poly(C) sequence (C 144 -C 148 ) was identified upstream of the stem-loop but within the AB region. A CCCU-CUCC motif (C 190 -C 197 ) was also identified 2 nt downstream of the stem loop (Fig. 1A). Another sequence with a poly(C) consensus element, C 102 -C 107 (Fig. 1A), resides outside the AB region but within the H-ferritin mRNA 5Ј-UTR. The presence of these poly(C) stretches raise the possibility that the PCBPs may bind the AB and the H-ferritin 5Ј-UTR and function to regulate translational efficiency, similar to that observed for 15 lipoxygenase (22) and poliovirus RNA (20) (for review. see Ref. 17).
To determine whether trans-acting factors were associated with the functionally active H-ferritin mRNA AB, REMSA was utilized. Two distinct RNA-protein complexes (RPC1 and -2) were identified in REMSA with a 32 P-labeled H-ferritin AB cRNA (33 nt) probe and HepG2 cytoplasmic extracts (see "Experimental Procedures"), both of which were competed for equally in a dose-dependent manner by excess unlabeled Hferritin AB (H-AB) cRNA (Fig. 7A, lanes 2-4). Similarly, a 200-fold excess of unlabeled L-ferritin AB (L-AB) also abolished detectable complexes with the probe (Fig. 7A, lane 5), suggesting that both H-and L-ferritin ABs have a similar affinity for HepG2 cell RNA-binding proteins. A 200-fold excess of unlabeled homologous cRNA of the acute phase reactant, AGP, did not compete for formation of RPC1 and -2 (Fig. 7A, lane 6), emphasizing the specificity of the interaction. Of interest, a 200-fold excess unlabeled ⌬3 and Mc cRNAs (Fig. 7A, lanes 7  and 8, respectively) did not compete as efficiently as unlabeled H-or L-ferritin AB cRNAs, consistent with a sequence and structure-specific requirement for the trans-acting factors. It is noteworthy that cold Mc RNA competed more efficiently than cold ⌬3 mRNA, consistent with the cis-regulatory data, demonstrating a sequence and structural requirement for optimal function of the AB as a translational enhancer (Fig. 6). An increase in complex formation was observed in the presence of a 200-fold excess of an unlabeled IRE cRNA (Fig. 7A, lane 9), suggesting that either there are important protein-protein interactions between iso-IRPs and proteins forming RPC1 and -2, or the IRPs compete with other RNA-binding proteins to bind to the AB.
UVXL-Western assay with a HepG2 cell cytoplasmic extract and the 32 P-labeled H-ferritin AB riboprobe (63 nt) demonstrated several RPCs of M r 43-68 (Fig. 7B, lane 1). These complexes were effectively competed by excess unlabeled AB cRNA (Fig. 7, B (lane 2) and C) but not by excess poly(A) (Fig.  7, B (lane 3) and C). However, excess poly(C) significantly reduced formation of the 68-, 48-, and 43-kDa complexes (Fig.  7, B (lane 4) and C). Varying concentrations of unlabeled tRNA (0 -1 g) had no effect on the RNA-protein complex (data not shown). The relative intensities of each of the complexes determined by ImageQuant analysis and plotted (Fig. 7C) highlights that the 43-kDa complex was reduced the most by competition with poly(C) RNA, consistent with the known M r of ␣CP1/␣CP2 at ϳ43 kDa (17). Probing the same membrane with ␣CP1 and ␣CP2 Abs confirmed that both proteins were present in the extracts and that excess competitor RNA did not effect their concentrations in samples (Supplemental Fig. 2A). Taken together, these data raised the possibility that the AB could be a target for binding by ␣CP1 and/or ␣CP.
We next examined the effects of IL-1␤ on RPC formation from HepG2 cells treated over a 16-h time course using UVXL-Western assays. IL-1␤ treatment decreased binding of the RPCs (43-48 kDa) over the 16-h period (by ϳ40% as determined by PhosphorImager-ImageQuant analysis), whereas actin levels remained constant (Fig. 7D). In addition, IL-1␤ had little or no effect on the binding interaction between IRE and IRP in hepatoma cells (data not shown). To determine whether treatment with IL-1␤ altered total cellular levels of ␣CP1 and ␣CP2 protein, we examined lysates from the cells from the 16-h IL-1␤ time course. We consistently observed that IL-1␤ did not change the intracellular levels of ␣CP1 and ␣CP2 in HepG2 cells (Supplemental Fig. 2B). These data indicated that IL-1␤ signals modulated binding to the AB region rather than altering absolute levels of ␣CP1 and ␣CP2.
To investigate the association of ␣CP1 and ␣CP2 with Hferritin mRNA in vivo and to investigate the effects of iron on this interaction, we utilized an IP-RT-PCR assay in HepG2 cells. H-ferritin mRNA could be detected in all HepG2 lysate supernatants (Fig. 7E, lanes 1-5), producing an amplicon of 290 bp, the size of the positive H-ferritin control (Fig. 7E, lane 17). ␣CP2 co-immunoprecipitated with H-ferritin mRNA in lysates from untreated HepG2 cells (Fig. 7E, lane 8), whereas association of H-ferritin mRNA with ␣CP1 was not detected (Fig. 7E,  lane 7). However, ␣CP1 preferentially associated with H-ferritin mRNA in lysates from HepG2 cells treated with 10 M iron (Fig. 7E, lane 9). Residual ␣CP2 binding remained under these conditions (Fig. 7E, lane 10). In the presence of IL-1␤ neither ␣CP1 nor ␣CP2 co-immunoprecipitated with H-ferritin mRNA (Fig. 7E, lanes [11][12] even in the presence of iron (Fig. 7E, lanes [13][14]. This is consistent with the data in Fig. 7D in which IL-1␤ reduced complexes at 43-48 kDa, which are predicted to contain one or more of the PCBPs. Western blotting confirmed that the HepG2 cells exhibited a consistent physiologic response to ligand; that is, an increase in both H-and L-ferritin protein levels in response to iron and IL-1␤, with an augmented response when added together (Fig. 7F). Taken together these data suggest that ␣CP1 and ␣CP2 both associate with H-ferritin mRNA in vivo and that the interaction is regulated by the prevailing intracellular iron concentration.
Given the IP-RT-PCR data demonstrating in vivo association of the ␣CPs with H-ferritin mRNA, we next sought to examine the binding between recombinant ␣CP1 and ferritin mRNA. In REMSA assays recombinant ␣CP1 bound avidly to the AB probe (Fig. 8A), whereas no significant binding was observed with high concentrations of GST protein alone. In UVXL assays, recombinant ␣CP1 binding to the H-ferritin AB probe was effectively competed with excess unlabeled H-ferritin AB and poly(C) mRNAs but not with poly(A) mRNA (Fig. 8B). Excess tRNA (1 g) did not compete with the ␣CP1-ferritin mRNA interaction (data not shown). An ␣CP1 Western blot confirmed that these changes were unrelated to alteration in the input concentration of ␣CP1 (Fig. 8B, right-hand  kDa). These data suggest a specific direct interaction of ␣CP1 with H-ferritin AB mRNA and support our sequence homology observations and IP-RT-PCR results. DISCUSSION These studies have defined the role of the H-ferritin AB cis-element within the 5Ј-UTR as a significant translational regulatory element and identified ␣CP1 and ␣CP2 as novel trans-acting factors that associate with the H-ferritin mRNA in vivo. The data indicate that the H-ferritin mRNA AB is a substantial base-line translation enhancer element of H-ferritin mRNA in addition to being IL-1␣ and IL-1␤ responsive in two separate hepatoma cell lines, HepG2 and Hep3B. Before this work, the only RNA-protein interaction controlling ferritin synthesis that had been well characterized was the regulated binding between the IRE RNA stem-loop and the iso-IRPs (IRP1 and IRP2) in response to intracellular iron chelation and oxidative stress (1)(2)(3)(4)(5)(6)(7)(8). However, the data presented in this report provide strong evidence that the 63-nt GC-rich AB sequence is a novel additional cis-element that functions together with the IRE to regulate ferritin expression. Moreover, the data demonstrate that the AB mediates IL-1␤ signaling more efficiently in Hep3B cells than in HepG2 cells, indicating that the translational enhancement exhibited in vivo by the AB is present in more than one cell line and is cell line-specific.
We found that the translation mediated by the AB was up-regulated by both IL-1␣ and IL-1␤ in a time course reflecting that of endogenous ferritin translation. The data in Fig. 2E showed that translation of new ferritin L and H subunits began 6 h after IL-1␤ treatment of HepG2 cells. Similarly, we found in our transfection experiments that IL-1 treatment induced translation of an H-ferritin 5Ј-UTR-driven CAT reporter at 6 h post-cytokine stimulation (Fig. 3A). In contrast to the stimulation with IL-1 treatment, however, IL-6 inhibited CAT translation driven via the AB (Fig. 6B). This ϳ10 -20% inhibition was modest but significant (p Ͻ 0.05, n ϭ 8) and was matched from the same data set by an average 2-fold increase in IL-1induced CAT translation, emphasizing the cytokine specificity for the translation enhancement mediated by the AB.
Our data demonstrated that the -fold increase of IL-1-dependent conferred CAT gene expression (ϳ2-3-fold) was always less than the ϳ4-fold increase in ferritin expression (compare Fig. 2, panel E, with Figs. 3 and 6 in HepG2 cells). This divergence between the IL-1-increased AB-dependent CAT expression and ferritin synthesis may result from additional IL-1 signaling through 3Ј-UTR sequences in the ferritin transcript (26,30,31). Subsequent investigation may demonstrate that IL-1 induces the H-ferritin mRNA 3Ј-UTR to interact with the AB as an additional component in the pathway required for full induction of ferritin expression. Indeed a cooperative interaction between the 3Ј-UTR and 5Ј-UTR was recently reported for IL-1␣-and IL-1␤-dependent and basal translation enhancement of the amyloid precursor protein transcript, another metalloprotein that is associated with Alzheimer disease (31).
Our finding that the sense version of the H-ferritin AB domain formed a stable stem-loop and was a base-line translational enhancer was unexpected since several groups have reported that similar stable RNA 2°structures normally repress translation of a downstream reporter (10,(32)(33)(34). Interestingly, the Gibbs free energy of the sense version of the Hferritin AB (pSV2(Ac)CAT) and the corresponding antisense version (pSV2(rev)CAT) are equivalent (⌬G ϭ Ϫ40 kcal/mol). However, the antisense version of the AB inhibited downstream CAT translation, consistent with a model implicating RNA impedance of the scanning 43 S ribosome (10, 34). There have been sporadic reports detailing positive regulation of translation determined by RNA structure as was displayed by the AB in this study. Park (35) showed that selective influenza viral mRNA translation is mediated by the cellular RNA-binding protein, GRSF-1, and Wulczyn and Kahmann (36) showed that the bacteriophase mu Com protein can be activated to specifically bind to the side of viral mom RNA during translation activation of viral synthesis. Fu et al. (37) identified another GC-rich cis-element with a stable 2°RNA structure that activated mRNA translation in fertilized oocytes. The transferrin mRNA 5Ј-UTR also mediates positive regulation of transferrin in response to intracellular iron chelation with DesF (38,39). In addition, the ferritin IRE stem-loop is a positive regulator of translation in the absence of IRPs (in rabbit reticulocyte and soybean extracts) to translate IREdriven ferritins (40), consistent with a requirement that the IRE is present for efficient translation of ferritin. Thus, the H-ferritin mRNA AB adds to an emerging list of translational enhancers characterized by a stable stem-loop structure that act selectively when folded in a wild-type structure.
To account for the possibility that the difference between the CAT translation induced with the wild-type H-ferritin (5Ј-UTRCAT) and the HIRECAT vectors was due to the altered size of the constructs, we generated a nonsense scrambled spacer (Scr-HIRECAT) of the same length as the wild-type AB. Our transfection studies confirmed that the Scr-HIRECAT vector was unresponsive to any of the cytokine treatments, confirming the specificity of the AB sequence for the translational enhancement. Thus, the wild-type AB element represents a bona fide translation enhancer in the natural context of the H-ferritin promoter. This predicted RNA 2°structure of the AB box in the HIRECAT construct (Fig. 3E) was identical to the AB RNA stem-loops predicted for H-ferritin mRNA 5Ј-UTR (Fig. 6) and the 5Ј-UTRCAT construct (Supplemental Fig. 1A). Significantly, no similar stem-loop was predicted to exist in the 5Ј-UTR of the chimeric H-ferritin/CAT 5Ј-UTR expressed from the Scr-HIRECAT construct (Fig. 3F). Taken together, these data provided compelling evidence that the RNA 2°structure of the AB represents a novel structure-specific translational control sequence in H-ferritin mRNA additional to the better characterized IRE.
There are increasing examples of mutations within the 5Ј-UTR sequence of a target mRNA that influence the normal function of the RNA, modify translational efficiency, and result in an important clinical phenotype. For example, hereditary thrombocythemia (familial essential thrombocythemia or familial thrombocytosis) is associated with mutations in upstream AUG codons in the 5Ј-UTR of the thrombopoietin mRNA that normally function as a translational repressor. Their inactivation leads to excessive production of thrombopoietin and elevated platelet counts (32). In the familial hyperferritinemia cataract syndrome several different point mutations or deletions in the 5Ј-UTR of the L-ferritin IRE have been identified. These alterations reduce the affinity of the IRPs for the IRE, which normally inhibit ferritin mRNA translation, and as a consequence lead to an increased production of Lferritin (33). Interestingly, each unique mutation confers a characteristic degree of hyperferritinemia and severity of cataract in affected individuals. Our mutation analysis confirmed that maintenance of the wild-type RNA structure in the Hferritin AB mRNA is crucial for efficient basal and IL-1␤-dependent translation of CAT reporter mRNA. It was of interest to note that HepG2 cells transfected with the ⌬3 mutant lacked IL-1␤-inducible CAT expression and also displayed a 5-10-fold reduction in base-line CAT activity compared with wild-type controls. Our structural predictions were further validated with data using the Mc mutant, which has a C to A substitution and is predicted not to alter the H-ferritin AB mRNA stem-loop configuration. The Mc mutant enhanced base-line reporter gene regulation and IL-1␤-induced translation in hepatoma Hep3B cells, although this effect was less than in HepG2 cells (Figs. 5 and 6). It will be of interest to determine whether natural mutations exist in the human ferritin AB domain and if so whether they are associated with a clinical phenotype reflective of aberrant ferritin homeostasis.
The PCBPs ␣CP1 and ␣CP2 are members of the heterogeneous nuclear ribonucleoprotein K-homology domain family of RNA-binding proteins (17) and regulate the expression of a variety of transcripts, including ␣-globin, tyrosine hydroxylase, and erythropoietin (23,(41)(42)(43)(44)(45) as well as regulating translation of 15-lipoxygenase (22) and human Papillomavirus (17,20). We have recently identified in vivo association of ␣CP1 with several transcripts including the human androgen receptor (18), p21 WAF1 (46), and renin (47). In the latter, ␣CP1 plays an important role in regulating the stability of renin mRNA. Identification of this diverse set of key mRNA targets for the CPs has focused attention on determining their functional role in human pathology.
Our in vivo association of ␣CP1 and ␣CP2 with H-ferritin mRNA from HepG2 cells suggests for the first time that these proteins bind to one or more of the motifs distributed throughout the H-ferritin mRNA (Fig. 1A) and that they may play an important role(s) in the regulation of ferritin gene expression. In particular, putative PCBP binding motifs in the AB region include a 5-base poly(C) stretch from ϩ144 to ϩ148 nt, a CCCTCTCC motif at ϩ190 to ϩ197 nt from the H-ferritin mRNA 5Ј cap site, and a further stretch from ϩ102 to ϩ107 nt. Our REMSA and UVXL studies confirmed that ␣CP1 binds the AB domain mRNA, consistent with this region being an important target within the 5Ј-UTR. Most interestingly, our data indicate that association of PCBPs with H-ferritin mRNA is an iron-and IL-1␤-dependent phenomenon. Thus, regulation of PCBP-ferritin mRNA interactions by intracellular iron and oxidative stress may be critical determinants of the overall ferritin subunit translational rate and protein synthesis.
Our data are consistent with a model whereby the AB is a docking site for 60 S ribosome subunit entry (48). Interestingly, DesF causes translational repression, which exerts a dominant effect over the IL-1␤ induction of ferritin subunit synthesis (13) (Fig. 2). This is because iron chelation by DesF suppresses ferritin translation at the first stage by inducing a high affinity interaction between IRP1/IRP2 and the 5Ј cap-specific IREs, which prevents IL-1␤ from exerting its effects. This suggests that IL-1␤-induced ferritin translation is dependent upon the pre-formation of a scanning complex, which is a requirement for establishing ferritin synthesis during 60 S ribosome entry 5Ј of the H-ferritin AUG codon (13).
According to the scanning model of translation the position of RNA stem-loops within the 5Ј-UTR is a critical determinant for translational repression of mRNAs (10,34,30). Goossen and Hentze (49) provided firm experimental support for a model in which ferritin is translated according to the "Kozak scanning model" (49). Ostareck et al. (22) showed that ␣CP1 is involved at inhibiting 60 S ribosome joining when binding to the 15lipoxygenase 3Ј-UTR (22). Reduced binding of ␣CP2 to the AB in the H-ferritin mRNA may likewise promote 60 S ribosome subunit joining and enhanced translation, although the mechanisms by which the PCBPs are involved in ferritin translation will require further study.
Binding by the IRPs may prevent the AB from forming an IRES and could explain its function in the context of ribosome scanning (50 -53). Consistent with this model, ␣CP2 is known to promote translation of poliovirus mRNAs after infection of HeLa cells by selectively interacting with domain IV of an IRES in the 5Ј-UTR of the virus RNA (20). However, many endogenous eukaryotic mRNAs, including the vascular endothelial growth factor, are translated from IRESs downstream from the 5Ј cap sites of their mRNAs (5Ј cap-independent translation) (54 -56). These data emphasize the need to evaluate the functional role of ␣CP1 and ␣CP2 in H-ferritin translation mediated via the AB domain.
Maintenance of efficient ferritin translation by the cytokineresponsive AB domain is consistent with the fact that ferritin provides cytoprotection to liver cells through stress-responsive pathways mediated both via cytokines (16) and accelerated by iron (heme)-catalyzed oxidative stress (57). Ferritin is at high abundance in the brain to protect neuronal cells from metal catalyzed oxidative damage during inflammation (9,58,59). Ferritin is also present in amyloid plaques of AD, suggesting a neuroprotective role in the final stages of neuronal death during this disease process (9,58). Translational activation through the ferritin AB domain also provides a mechanism for iron sequestration from the bloodstream into liver and spleen as occurs in rat models in the anemia of chronic disease and inflammation (16,60). Taken together, these data support a model for the maintenance of high basal ferritin levels by dual control of ferritin mRNA translation in the presence of intracellular iron through the upstream IRE and the downstream AB domain. Certainly regulation of ferritin mRNA by 5Ј-UTRspecific AB domains remains key to our understanding of the regulation of ferritin expression during inflammation, where it plays a role as the universal iron storage protein that protects cells from metal-catalyzed oxidative stress (16,57,61). The studies described here provide novel insight into the function of the AB and lay the foundation for future work that will investigate the precise functional role of the AB. It will be of interest to determine whether the AB works in a cooperative manner with the IRE and to delineate how the PCBPs integrate with the IRPs to coordinately regulate iron homeostasis in cells under a variety of clinical states.